Research On RFID Data Intelligent Cleaning Method In Manufacturing Workshop

With the deep combine of contemporary information technology and manufacturing,manufacturing mode of manufacturing enterprises is being changed by intelligent facility,intelligent factory based on Cyber-Physical System.Radio Frequency Identification,which is a fast and convenient and non-contact collection and processing information technology,can perceive information resource dynamically such as monitoring production and processing status,optimizing control of logistics and warehouse,intelligent tracking and tracing,which can make manufacturing workshop more and more digital and intelligent.So RFID equipment is more widely used in the manufacturing plant.However,because of complexity and dynamicity of RFID,interference of electromagnetic wave,reflection of metal and complexity of perceived object in manufacturing workshop,raw RFID readings are massive,multi-source,high noise and high temporal and spatial correlation.The development of RFID in the field of intelligent manufacturing is being restricted by the unreliability of RFID readings,this paper analyzes the characteristics of RFID data streams and the existing data problems in manufacturing plant.In view of the existing problems,this paper proposes improved methods and establishes a comprehensive data cleaning model and a cleaning system.The main work of this paper is as follows:(1)The characteristics of RFID data and the reason of false negative readings,false positive readings and duplicate readings in manufacturing plant are deeply studied.Combined with the characteristic of different RFID topologies with different data problems,an integrated data-oriented division cleaning architecture is established.Firstly,RFID data is divided into duplicate readings and dropped readings based on the data characteristics,secondly,duplicate readings are read into the spatial redundancy processing module and dropped readings are read into the leakage processing module,finally the integrated data is processed by temporal redundancy module.(2)In view of the dropped readings,the paper considers the tags' transition,dynamically adjusts parameter ? and utilize median filter to eliminate isolated noise in order to further optimize the existing WSTD algorithm.In view of the duplicate readings,based on Bayesian inference,because of the limitations of the existing n-state detection model and two-dimensional detection model on the scene,a universal and robust detection method based on spatial relationship model is introduced,which makes wider application range of this algorithm and enhances efficiency.(3)Extensive simulations in MATLAB are carried out to analyze the advantages and disadvantages and to verify the accurateness and effectiveness of the proposed algorithms over classical algorithms and improved methods of classical algorithm.The paper develops a RFID data cleaning system based on Java Web to verify the availability of the proposed comprehensive cleaning model and cleaning method in practice.